Meet Claude

Claude is a large language model (LLM): an AI trained on an enormous amount of text to do one deceptively simple thing extremely well - predict what words should come next, given everything it has read so far. Out of that single ability comes the whole range of what you experience: answering questions, summarising, drafting, translating, reasoning step by step, writing code. It is worth holding onto that mechanism from day one, because it explains both Claude's brilliance and its failures. Claude produces the most plausible continuation of a conversation. Usually the most plausible answer is also the correct one - which is why it is so useful. But not always, and when a confident, fluent, plausible answer happens to be wrong, nothing in the machine flags it. Lesson 0.2 goes deeper into this; for now, the one-line version is: Claude is a plausibility engine, not a truth engine, and you supply the truth-checking.

You reach Claude in a few ways. The everyday one is Claude.ai - a chat window in your browser, on the desktop app, or on your phone, where you type, paste, and attach files. Anthropic also offers Claude Code (a version that lives in a developer's terminal for writing software), and an API that lets other software - including tools built into Studio Matrx - call Claude behind the scenes. For most of this course we work in Claude.ai, because that is where an architect or interior designer does ninety per cent of their work, and it needs no technical setup at all.

Two features inside Claude.ai will recur throughout the course, so meet them now. Projects are persistent workspaces: you give a Project background knowledge (your practice's standards, a client brief, a materials palette) and custom instructions once, and every chat inside it starts already knowing them - the difference between briefing a new assistant every single morning and keeping one who remembers. Artifacts are live documents that appear beside the chat - a spec, a schedule, a checklist, even a small working web tool - that Claude writes and revises in place while you watch, instead of burying the result in a wall of chat text. Modules 1 and 9 make both of these part of how you actually work.

The single most useful mental model for this whole course is not 'Claude the oracle' but Claude the brilliant intern - phenomenally fast, widely read, never tired, happy to redo something ten times, and genuinely helpful - who is nonetheless an intern: not the person who signs the drawings, meets the client's eye, or carries the can when something is wrong. You would never let an intern's first draft go out under your seal unread. You direct them, you read what comes back, you correct it, and you own the result. Claude is exactly the same, minus the coffee breaks.

That relationship has a shape, and it is worth naming because every good workflow in this course is a version of it: you direct (frame the task, give context, set the format and constraints); Claude drafts (a summary, options, an analysis, a draft); you judge (is this true, useful, on-brief, defensible?); and you refine (re-prompt, pick and combine, correct, or throw it away) - then round again. Claude never closes that loop by itself. Remove yourself from it and you do not have an efficient office; you have unread output going out under your name. Three of those four steps are yours - which is exactly why a trained designer, not a prompt hobbyist, gets the most out of Claude.

How hard you judge should scale with what is at stake. A moodboard word-list or a brainstorm of concept directions is low-stakes and divergent - you were going to curate ruthlessly anyway, so let Claude be wild and plentiful and check it lightly. A fire-egress clause, a structural figure, a specification that goes to site is high-stakes and convergent - here a confident-wrong answer is a real liability, and you verify every claim against the actual source. Matching your scepticism to the stakes, rather than trusting or distrusting Claude uniformly, is the core professional discipline, and we return to it in every module.

Claude is not one trick for one stage - it shows up across the entire arc of a project, which is why this course follows that arc rather than a feature list. At the front of the job it researches precedents, navigates codes and byelaws, and reads site and climate context (Module 2), then helps turn a messy client conversation into a structured brief, space program and adjacency logic (Module 3). In the design phase it is a foil for developing and articulating a concept, and for writing a rationale that actually holds together (Module 4). Through delivery it drafts the correspondence a practice runs on - proposals, fee letters, emails, minutes, reports (Module 5) - and the documentation it is quietly excellent at: specifications, schedules, BOQ and drawing-note text, always with you as checker (Module 6). It supports decisions by laying out options, trade-offs, rough cost and performance reasoning (Module 7), and it can even act as a coding partner for designers who do not code, scripting Grasshopper, Revit and small tools in plain-English instructions (Module 8).

Notice what is common to all of these: in every case Claude does the fast, voluminous, first-draft, or look-it-up work, and you do the deciding. It is not replacing the architect at any of these stages; it is removing the drudgery that sits between you and the judgement only you can make. Used this way across a week, the time it gives back is substantial - and it goes straight into design, into clients, and into getting home earlier.

One honest caution to carry from the very start: Claude renders plausible prose, numbers and code, not verified fact. A specification clause it drafts may cite a standard that does not say quite what it claims; a cost it estimates is a sanity-check, not a quantity survey; a byelaw summary is a lead to confirm, not a ruling. This is not a reason to avoid it - it is a reason to use it the way a good principal uses a junior: gratefully, and with a red pen always in hand.

This is a practical, practice-first course, not a tour of demos. Over eleven modules you will first learn to talk to Claude well - prompting for design, giving it lasting context with Projects, building live documents with Artifacts, and working from your own drawings and PDFs (Module 1). Then you will apply it across the arc: research and discovery (2), the brief and programming (3), concept and narrative (4), writing the practice (5), specifications and documentation (6), and analysis and decision support (7). Module 8 turns Claude into a gentle coding and automation partner; Module 9 shows how to run it as a studio system across a team; and Module 10 tackles the non-negotiables - confidentiality, client data, accuracy, liability, authorship, IP and disclosure - and what a lasting, Claude-augmented practice looks like.

Two things stay constant throughout. First, the emphasis is on judgement over tricks: models and buttons change every few months, but the discipline of directing well and checking hard does not, and that is where your design training gives you an edge a prompt hobbyist will never have. Second, everything is taught for real practice in the real world - Indian codes, climates and clients woven in alongside international ones, and honesty about cost, limits and hype rather than salesmanship. Studio Matrx is free and not-for-profit, and this course, like the rest of the Academy, exists to make you genuinely better at your work - not to sell you a tool. Sister courses go deeper where this one stays Claude-specific: AI-Assisted Design Workflows for the whole AI toolscape, Generative AI for Architecture & Interiors for image generation. Take this one to make Claude a dependable part of how you practise.